[AArch64][GlobalISel] Select immediate fcmp if the zero is on the LHS.

llvm · Jan 15, 2021 · aa8a2d8 · aa8a2d8
1 parent ceaf011
commit aa8a2d8
Show file tree

Hide file tree

Showing 2 changed files with 35 additions and 0 deletions.
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
@@ -4224,6 +4224,14 @@ AArch64InstructionSelector::emitFPCompare(Register LHS, Register RHS,
   // to explicitly materialize a constant.
   const ConstantFP *FPImm = getConstantFPVRegVal(RHS, MRI);
   bool ShouldUseImm = FPImm && (FPImm->isZero() && !FPImm->isNegative());
+  if (!ShouldUseImm) {
+    // Try commutating the operands.
+    const ConstantFP *LHSImm = getConstantFPVRegVal(LHS, MRI);
+    if (LHSImm && (LHSImm->isZero() && !LHSImm->isNegative())) {
+      ShouldUseImm = true;
+      std::swap(LHS, RHS);
+    }
+  }
   unsigned CmpOpcTbl[2][2] = {{AArch64::FCMPSrr, AArch64::FCMPDrr},
                               {AArch64::FCMPSri, AArch64::FCMPDri}};
   unsigned CmpOpc = CmpOpcTbl[ShouldUseImm][OpSize == 64];

diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/select-fcmp.mir b/llvm/test/CodeGen/AArch64/GlobalISel/select-fcmp.mir
@@ -107,3 +107,30 @@ body:             |
     %3:gpr(s32) = G_FCMP floatpred(oeq), %0(s64), %2
     $s0 = COPY %3(s32)
     RET_ReallyLR implicit $s0
+...
+
+---
+name:            zero_lhs
+alignment:       4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0, $s1
+
+    ; CHECK-LABEL: name: zero_lhs
+    ; CHECK: liveins: $s0, $s1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr32 = COPY $s0
+    ; CHECK: FCMPSri [[COPY]], implicit-def $nzcv
+    ; CHECK: [[CSINCWr:%[0-9]+]]:gpr32 = CSINCWr $wzr, $wzr, 1, implicit $nzcv
+    ; CHECK: $s0 = COPY [[CSINCWr]]
+    ; CHECK: RET_ReallyLR implicit $s0
+    %0:fpr(s32) = COPY $s0
+    %1:fpr(s32) = COPY $s1
+    %2:fpr(s32) = G_FCONSTANT float 0.000000e+00
+    %3:gpr(s32) = G_FCMP floatpred(oeq), %2(s32), %0
+    $s0 = COPY %3(s32)
+    RET_ReallyLR implicit $s0
+
+...